HEC-HMS Version 4.13

By: David Ho, P.H.

The Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS) software is designed to simulate precipitation-runoff processes of watershed systems. HEC-HMS provides a wide range of scalable methods for modeling hydrologic processes, delivers a modern and efficient user interface, supports robust optimization and uncertainty analysis capabilities, and has complete documentation and training options publicly available to the engineering community for free.  HEC-HMS is designed to be applicable for a wide range of geographic areas to solve the widest possible range of problems. This includes large river basin water supply and flood hydrology, and small urban or natural watershed runoff. 

The HEC-HMS team works hard to meet the modeling needs of USACE, as well as those of local, state, and other federal partners. Recent advances—such as improved support for gridded inputs, stronger GIS capabilities, faster multi-iteration uncertainty analyses, and expanded optimization features—have helped keep HEC-HMS one of the most widely used hydrologic software packages worldwide.  Additionally, the HEC-HMS team is committed to keeping the documentation current with the latest features and bug fixes.

The current development release of HEC-HMS is version 4.13. The development cycle included six beta releases before the final release in July 2025. Version 4.13 delivers numerous updates, including a new approach for specifying gridded parameters, a generalized framework for importing gridded data, enhancements to the Frequency Storm Meteorologic Method, and performance improvements for Ensemble Analysis. Several of these new features are highlighted below.

Specified Values Gridded Parameterization

In previous versions, gridded methods, e.g. Gridded Deficit & Constant, were initialized exclusively from state grids in the form of HEC-DSS file and pathname references. State grids are gridded datasets that store the model’s per-cell state (parameter values) at the start of a simulation, such as soil moisture/deficit, storage, or other variables. Users had to pre-generate these grids outside of HMS in order to use them. The specified values gridded parameterization feature adds the capability to parameterize values for each grid cell using the "Specified Values" initialization type (Figure 1).  Additionally, when using the "State Grids" initialization type, gridded references from HEC-DSS, ASC and GeoTIFF data sources can be used to initialize state.

This feature enables true distributed parameterization that can be directly specified by the user rather than relying on a single "lumped" value. Because each cell can be uniquely specified, you can represent spatial heterogeneity explicitly at the resolution of the grid—differences in soil water deficit, storage, or other variables such as transform parameters, canopy parameters, and baseflow parameters. This enables a more realistic, distributed representation of runoff generation across highly heterogeneous watersheds (Figure 2).  This can improve model results by better matching the spatial pattern and timing of runoff contributions, reducing bias and timing errors in simulated hydrographs when heterogeneity is an important driver of the response.

Parallelizing the Uncertainty Analysis Compute Type

Uncertainty Analysis computations typically require many iterations to adequately sample parameter distributions and characterize the range of plausible hydrologic outcomes. In earlier versions, these runs were limited to a single core, which could make large ensembles prohibitively slow. Starting in version 4.13, Uncertainty Analysis can leverage the set of multiple cores now typically available in modern computers to run simulations in parallel, substantially reducing overall runtime. Testing has shown runtimes are reduced by roughly 50 percent (Table 1), making it more practical to run larger ensembles, explore more parameters, and complete uncertainty-driven studies within typical project schedules.

This improvement is especially important for workflows that depend on thousands of simulations, such as Stochastic Storm Transposition (SST). SST repeatedly transposes precipitation grids to generate many plausible storm placements within a transposition region and then routes each realization through the hydrologic model to build flood-frequency information. Parallel execution helps users model a large enough ensemble to produce stable frequency estimates and meaningfully represent variability due to storm location, watershed heterogeneity, and nonlinear runoff processes without having to simplify the analysis solely to reduce compute time.

Table 1: Uncertainty Analysis runtime comparison between version 4.12 vs 4.13

Testing results performed on a 64-bit Intel Xeon W-10885 CPUY at 2.40 GHz, 32 GB RAM, and 8 Cores. The test project used Specified Values Monte Carlo method sampling from 3 storm parameters.

Generalized Gridded Data Reader Framework - NetCDF Support

A new, modernized framework was implemented for reading gridded data in HEC-HMS. This new architecture is important because it reduces reliance on a single gridded data container (HEC-DSS) and makes it easier to incorporate the kinds of datasets hydrologists increasingly use in practice.

The framework supports a broader range of gridded data formats, including NetCDF (Figure 3), which is widely used for large, multi-dimensional scientific datasets such as gridded meteorology, climate reanalysis, forecast model outputs, and other time-varying raster products. By supporting additional formats, HEC-HMS can more directly consume externally produced datasets without requiring extensive conversion workflows, which can introduce errors, add processing time, and complicate model reproducibility.

In addition, the framework includes intelligent caching strategies that improve read performance while managing memory usage efficiently. This matters for hydrologic modeling because gridded simulations often require repeatedly reading many timesteps of large rasters (sometimes for multiple parameters or multiple realizations). Faster, more efficient I/O reduces total runtime, improves interactive workflows, and makes large ensemble analyses more feasible.

Overall, this new gridded-data architecture enhances robustness (fewer workflow failure points tied to file handling and format constraints) and flexibility (support for current datasets and a clear path to incorporate additional gridded data sources in the future), helping users integrate a wider variety of observational, forecast, and climate-derived inputs into HMS modeling studies.

Modified Kling-Gupta Efficiency as Objective Function in Optimization Trial and as Calibration Statistic

The Modified Kling-Gupta Efficiency (MKGE) (Kling et al., 2012) was added as an Objective Function in the Optimization Trial and as a calibration statistic (Figure 4). The original Kling-Gupta Efficiency (KGE) was introduced by Gupta et al. (2009) to provide a more diagnostic, multi-component alternative to single-criterion measures such as mean squared error and the Nash-Sutcliffe Efficiency (NSE).

MKGE is recommended because it evaluates model performance using three complementary aspects of the simulated hydrograph:

• Correlation coefficient (r): how well the simulation matches the timing and shape (co-variation) of observed flows
• Bias ratio (β): how well the simulation matches the overall volume (systematic over- or under-prediction)
• Variability ratio (γ): how well the simulation matches the spread/variability of flows (for example, whether peaks and recessions are too “flashy” or too damped)

A reason MKGE is often preferred for calibration is that it makes tradeoffs explicit: a model should not be considered “good” if it gets the timing right but is strongly biased in volume, or if it matches volume but consistently underestimates variability and peak dynamics. In other words, MKGE discourages solutions that “win” on one aspect of fit while masking deficiencies in others—an issue that can occur with NSE or MSE, which tend to overweight high flows and can be difficult to interpret diagnostically.

MKGE can be decomposed into the three terms (r, β, γ). The MKGE value represents the most limiting of these components, which helps identify the dominant weakness controlling the overall score. The values of correlation, bias ratio, and variability ratio can be reviewed in the *.results files, enabling more targeted calibration (for example, adjusting loss parameters to address bias/volume, or transform/baseflow parameters to address variability and recession behavior).  Additional information is provided in the User's Manual and the Technical Reference Manual.

Frequency Storm Enhancements

Several enhancements have been made to the Frequency Storm meteorological model to improve usability, reduce setup burden, and produce more consistent storm hyetograph shapes.

• The “Annual-Partial Conversion” combo box has been removed from the user interface to simplify configuration and reduce the chance of selecting an incorrect conversion option. For projects created prior to HEC-HMS 4.13 that had “Partial to Annual” selected, rainfall depths are automatically converted during upgrade to 4.13, helping preserve backward compatibility and minimizing manual rework when migrating existing studies.

• The Frequency Storm method no longer requires that all frequency depths be explicitly provided in order to simulate a storm (Figure 5). Only the storm intensity depths and storm duration depths are required; any missing frequency depths are log-interpolated automatically. This is important because hydrologic frequency inputs are not always available at every duration (or may come from mixed sources), and requiring a complete set can force users to spend time assembling, estimating, or reformatting data purely to satisfy software requirements. By filling gaps internally using a consistent approach, HEC-HMS increases user flexibility, speeds up model setup, and makes it easier to run sensitivity studies across storm durations and return periods.

• A new “Re-sort Symmetrically” feature was added that re-sorts the rainfall depths into a pyramid-shaped pattern, where incremental depths increase toward the peak and then decrease afterward. In practice, users often expect a frequency storm to have a single dominant peak with a physically reasonable build-up and recession. If incremental depths are out of order, the storm pattern can look unrealistic (for example, multiple sharp peaks or large bursts late in the storm), which can lead to results that are difficult to explain or inconsistent with the intended storm temporal distribution. The symmetric re-sort option helps users quickly produce a storm pattern that aligns with common design storm assumptions, improves interpretability, and reduces the manual effort required to “fix” storm shape inputs.

Together, these updates make Frequency Storm modeling more streamlined and forgiving (fewer required inputs and UI decisions), while also supporting more defensible and consistent storm patterns that better match typical hydrologic design expectations.

Ensemble Analysis Performance Enhancements

Previously, ensemble analyses could experience noticeable slowdowns when opening or closing the Output Control dialog (Figure 6), as well as during the initial setup phase at the start of an ensemble compute. In HEC-HMS v4.13, significant performance improvements were implemented to reduce these delays, with the largest benefits seen in ensemble studies that include multiple large basin models.

These enhancements improve day-to-day usability by making it faster to configure and review output settings without lengthy UI waits, and they shorten “startup” overhead so more time is spent actually running simulations. This is especially useful for large ensemble workflows—such as sensitivity testing, and other multi-alternative studies—where users may repeatedly adjust settings and run many computes. The result is a more responsive experience and a more practical workflow for building and executing large ensemble analyses on typical project schedules.

Add Specified Values (Same Index) method to Monthly distributions

This feature enhances Uncertainty Analysis using the Monthly Distribution method by adding a new distribution type: Specified Values – Random (Same Index). With this option, users define paired data tables that list candidate values for each month (for example, a set of January values, a set of February values, etc.) for multiple parameters (Figure 7).

When two or more parameters are configured to use Specified Values – Random (Same Index), HEC-HMS will draw a single random index for a given month and apply that same index across all of the linked parameters’ tables. This ensures the sampled parameter values remain aligned (that is, they represent the same “realization” or “scenario” across parameters) rather than being combined randomly in ways that may be inconsistent or physically unlikely.

This matters because many hydrologic parameters are seasonally correlated or are calibrated/estimated as matched sets. For example, a “wet winter” realization might reasonably pair higher antecedent moisture-related values with higher groundwater/baseflow-related values, while a “dry summer” realization might pair a different consistent combination. The Same Index approach lets users preserve those intended relationships while still sampling uncertainty month-by-month.

At the same time, the method remains flexible: other parameters in the analysis can continue to use independent distribution types (normal, log-normal, uniform, triangular, other specified-value approaches, etc.) without being forced into the same-index behavior. This allows users to mix correlated monthly sampling where it is important with independent sampling where parameters are not expected to move together.

These are just a few of the new features added to HEC-HMS version 4.13.  Other improvements and features released with 4.13 include: 

• Modified Sorptivity-parameterized Green & Ampt Loss Method

• Grid to Point Additional Statistical Time Series
• User Specified Solar Shading Radius
• Changing ET Methods for Meteorological Model from the Parameters Menu
• Uncertainty Analysis for Subbasin Erosion
• Interpreting Additional Units for Initial Snow Depth and Snow Water Equivalent Grids
• Add Annual Pattern Option for Time of Max Daily Temperature
• Add Specified Values (Same Index) method to Monthly Distributions
• Option to include characteristics with GIS feature export

We encourage users to visit the HEC-HMS webpage to learn more about current software updates: https://www.hec.usace.army.mil/confluence/hmsdocs/hmsum/4.14/release-notes/v-4-13-0-release-notes

References

• Gupta, H. V., Kling, H., Yilmaz, K. K., & Martinez, G. F. (2009). Decomposition of the mean squared error and NSE performance criteria: Implications for improving hydrological modelling. Journal of Hydrology, 377(1–2), 80–91. https://doi.org/10.1016/j.jhydrol.2009.08.003

• Kling, H., Fuchs, M., & Paulin, M. (2012). Runoff conditions in the upper Danube basin under an ensemble of climate change scenarios. Journal of Hydrology, 424–425, 264–277. https://doi.org/10.1016/j.jhydrol.2012.01.011